The development and performance of charge-transfer potentials for water and silica sol-gel systems is considered. Charge-transfer potentials allow for redistribution of charge between atoms and therefore can describe their behavior with changes in local chemical environment; such behavior is critical to correct description of, for instance, the dissociation of water into hydroxide and hydronium ions. Parameterization of a complete potential using both charge-transfer functions and traditional two- and three-body empirical potential terms is a challenging numerical problem. Evolutionary Strategy (ES) optimizations were used to fit various charge-transfer potentials against first-principles data. Evolutionary strategies are similar to genetic algorithms in that they use parameter recombination and mutation in many simultaneous trial solutions with a competitive, generation-based search. Since there are many possible ES algorithms that could be used for such work, we first evaluated the efficiency of various evolutionary strategies in fitting two- and three-body empirical potentials for aqueous silicate systems against a ``training set'' of configurational energies. Finally, the performance of charge-transfer potentials in modeling pure water and the initial steps of silicic acid gelation is compared with that of the Feuston-Garofalini potential [J. Phys. Chem., 94, 5351 (1990)] and other recent models.